In this paper, we show that native speakers spontaneously divide the complex meaning of a new word into a presuppositional component and an assertive component. These results argue for the existence of a productive triggering algorithm for presuppositions, one that is not based on alternative lexical items nor on contextual salience. On a methodological level, the proposed learning paradigm can be used to test further theories concerned with the interaction of lexical properties and conceptual biases.